Biology Atlas

Map / Open register

question · deep · partly written

Do epigenetic clocks measure biology or assay noise?

Replicate noise moves a clock reading by up to 9 years, and published clocks disagree on reprogrammed cells, so the size and even the direction of a reported rejuvenation depend on the clock.

The short version

Replicate noise alone shifts six clocks by up to 9 years. On reprogrammed cells, published clocks gave about 40 years younger to about 13 years older. The size of any rejuvenation depends on the clock; none is shown to be wrong. 12

Reported, not yet replicated

The evidence more

Higgins-Chen and colleagues report technical-noise deviations up to 9 years between replicates for six clocks and a principal-component method that brings most replicates within 1.5 years; the author list includes people who built clocks. Kriukov and colleagues re-analysed public reprogramming data and report epistemic uncertainty: published clocks inconsistent, some zero or negative rejuvenation, and about 86% and 69% of CpGs rejected by a covariate-shift test in the two data sets compared with ageing skin. Gill and colleagues report a transcriptome rejuvenated by around 30 years on a novel transcriptome clock and a methylation clock rejuvenated to a similar extent. 123

Reported, not yet replicated

What is not claimed deep

None of these sources shows that clocks measure causes of ageing. The re-analysis is a re-analysis, not a wet-lab replication, and does not rule out age reversal. 2

Reported, not yet replicated

Observation

Technical noise produced deviations of up to 9 years between replicates for six prominent epigenetic clocks. A re-analysis of public reprogramming data found published clocks gave answers from about 40 years younger (Horvath multi-tissue) to about 13 years older (Hannum blood) at the same time point. A skin-trained model found significant rejuvenation between day 0 and day 17 (P = 0.014) with a day-17 credible interval of about 0 to 70 years. 12

Reported, not yet replicated

Hypotheses

Clock readings are dominated by technical noise

Replicate noise at single CpGs propagates into clock outputs; a principal-component version reduces it to within 1.5 years for most replicates. 1

Predicts. Replicates of the same sample differ by years in standard clocks and agree within about 1.5 years in principal-component versions.

Would falsify. Standard clocks agreeing within a small margin across replicates.

Reported, not yet replicated

Clocks are out of distribution on reprogramming data

Reprogramming methylomes are poorly represented in the ageing data clocks were trained on, giving high uncertainty and inconsistent predictions. 2

Predicts. Different clocks disagree on reprogrammed cells; covariate-shift tests reject many CpGs.

Would falsify. Clocks agreeing and showing small uncertainty on reprogramming data.

Reported, not yet replicated

Some rejuvenation is real

Maturation-phase transient reprogramming rejuvenated transcriptome and methylation clocks by around 30 years; the re-analysis does not question the possibility of age reversal. 32

Predicts. Rejuvenation visible on several independent measures and replicated by other labs.

Would falsify. Independent replication showing no change on clock-independent measures.

Contested

What is missing

A ground truth for biological age

The true biological age of reprogrammed cells is unknown, so clock predictions cannot be verified against it. 2

Reported, not yet replicated

Independent wet-lab replication

The opened sources include a re-analysis of public data but no independent replication of the reprogramming experiment. 23

Reported, not yet replicated

What is known

Clocks are statistical models; replicate noise and training-set mismatch both change their output. 12

Reported, not yet replicated

What is not

Whether a clock change in cultured cells reflects reversal of ageing biology; whether methylation drives ageing or only reflects it is not addressed here. 23

Reported, not yet replicated

What would settle it

Replicated experiments with several noise-corrected clocks plus non-clock measures of function, with the uncertainty of each reported. 12

Our inference

Who works on it

Still open here

Linked from

Sources

  1. Higgins-Chen, A.T., Thrush, K.L., Wang, Y., et al., Levine, M.E.. A computational solution for bolstering reliability of epigenetic clocks: Implications for clinical trials and longitudinal tracking. Nature Aging (2022). paper; opened 2026-11-17 all records citing this
  2. Kriukov, D., Kuzmina, E., Efimov, E., et al., Khrameeva, E.E.. Epistemic uncertainty challenges aging clock reliability in predicting rejuvenation effects. Aging Cell (2024). paper; opened 2026-11-17 all records citing this
  3. Gill, D., Parry, A., Santos, F., et al., Reik, W.. Multi-omic rejuvenation of human cells by maturation phase transient reprogramming. eLife (2022). paper; opened 2026-11-17 all records citing this
Where it sits on the map
History of this record (1)

This record has been changed 1 time, on the dates below. The note kept against each change is written for the people who maintain the atlas, in their own shorthand, so it is not shown here; it is in the record itself.

Kind
big-open
Register
open-question
Status
partly written
Last updated
2026-11-17
Record
content/questions/do-epigenetic-clocks-measure-biology-or-assay-noise.json